Body
An overview of the Claude models available in Claude Enterprise, with recommendations for which model and effort level to use for different tasks.
Overview
Claude Enterprise lets you switch between several Claude models and adjust how much effort each model puts into a response. The model you pick affects response quality, speed, and how many tokens you use. This article covers the available models and gives starting recommendations for common tasks.
The model selector sits at the bottom of the chat window, next to the message box. Select the current model name to open the picker, then choose More models to see the full list. Use the Effort control in the same menu to set how hard the model works on each response.
Helpful Resources
Assumption
You are a technician with a Claude Enterprise license and access to the model picker in the Claude chat interface.
Available models
The picker groups models into families. From most to least capable, they rank as follows:
- Fable
- Opus
- Sonnet
- Haiku
Several versions of each family may appear in the list, for example multiple Opus and Sonnet releases. Higher-ranked models generally handle harder problems better, but they can be slower and use more tokens.
Effort levels
Each model has an effort setting that controls how much work it puts into a response. A higher effort level tends to improve results on complex tasks at the cost of speed and tokens. You can change the effort level when you switch models.
Recommendations
These are practical starting points. Adjust them based on your own results.
General enterprise work
Use Opus at High effort. For complex agentic coding and enterprise work.
Coding or Implementing Structured Tasks
Use Sonnet at High effort. Great balance of speed and intelligence. Excels when given structured plans to execute.
System or Complex Business Process Design
Fable is the most capable family, but it consumes tokens quickly, so it is not recommended for routine use. Reserve it for tasks that specifically need its extra capability like deep system design or reviewing complex business processes. Next-generation intelligence for long-running agents.
Try it out
Experiment with the different models and effort levels to find what fits your work. The best choice depends on the task, how fast you need a response, and how many tokens you want to spend.